Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -5,6 +5,7 @@ import difflib
|
|
5 |
import tempfile
|
6 |
import os
|
7 |
import speech_recognition as sr
|
|
|
8 |
|
9 |
# Function to play the text (optional)
|
10 |
def play_text(text):
|
@@ -14,24 +15,26 @@ def play_text(text):
|
|
14 |
os.system(f"start {temp_file.name}") # Windows
|
15 |
return "✅ Text is being read out. Please listen and read it yourself."
|
16 |
|
|
|
|
|
|
|
17 |
def transcribe_audio(audio, original_text):
|
18 |
-
recognizer = sr.Recognizer()
|
19 |
try:
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
# Try chunking if needed
|
24 |
-
transcription = recognizer.recognize_google(audio_data, language="hi-IN")
|
25 |
|
26 |
-
|
|
|
|
|
27 |
import re
|
28 |
original_words = re.findall(r'\w+', original_text.strip())
|
29 |
transcribed_words = re.findall(r'\w+', transcription.strip())
|
30 |
|
31 |
matcher = difflib.SequenceMatcher(None, original_words, transcribed_words)
|
32 |
accuracy = round(matcher.ratio() * 100, 2)
|
33 |
-
|
34 |
-
|
|
|
35 |
|
36 |
result = {
|
37 |
"📝 Transcribed Text": transcription,
|
@@ -39,10 +42,6 @@ def transcribe_audio(audio, original_text):
|
|
39 |
"⏱️ Speaking Speed (words/sec)": speed
|
40 |
}
|
41 |
return result
|
42 |
-
except sr.UnknownValueError:
|
43 |
-
return {"error": "Could not understand audio"}
|
44 |
-
except sr.RequestError as e:
|
45 |
-
return {"error": f"Request error: {e}"}
|
46 |
except Exception as e:
|
47 |
return {"error": str(e)}
|
48 |
|
|
|
5 |
import tempfile
|
6 |
import os
|
7 |
import speech_recognition as sr
|
8 |
+
from faster_whisper import WhisperModel
|
9 |
|
10 |
# Function to play the text (optional)
|
11 |
def play_text(text):
|
|
|
15 |
os.system(f"start {temp_file.name}") # Windows
|
16 |
return "✅ Text is being read out. Please listen and read it yourself."
|
17 |
|
18 |
+
# Load model once (outside function for efficiency)
|
19 |
+
model = WhisperModel("small", compute_type="float32") # Or "medium" for better accuracy
|
20 |
+
|
21 |
def transcribe_audio(audio, original_text):
|
|
|
22 |
try:
|
23 |
+
# Run inference
|
24 |
+
segments, info = model.transcribe(audio, language='hi')
|
|
|
|
|
|
|
25 |
|
26 |
+
transcription = " ".join([segment.text for segment in segments])
|
27 |
+
|
28 |
+
# Clean and split the text better
|
29 |
import re
|
30 |
original_words = re.findall(r'\w+', original_text.strip())
|
31 |
transcribed_words = re.findall(r'\w+', transcription.strip())
|
32 |
|
33 |
matcher = difflib.SequenceMatcher(None, original_words, transcribed_words)
|
34 |
accuracy = round(matcher.ratio() * 100, 2)
|
35 |
+
|
36 |
+
# Speaking speed (approximate)
|
37 |
+
speed = round(len(transcribed_words) / info.duration, 2)
|
38 |
|
39 |
result = {
|
40 |
"📝 Transcribed Text": transcription,
|
|
|
42 |
"⏱️ Speaking Speed (words/sec)": speed
|
43 |
}
|
44 |
return result
|
|
|
|
|
|
|
|
|
45 |
except Exception as e:
|
46 |
return {"error": str(e)}
|
47 |
|